Cell type–specific plasticity in synaptic, intrinsic, and sound response properties of deep-layer cortical neurons after noise trauma

Abstract

Peripheral damage drives auditory cortex (ACtx) plasticity, but the underlying synaptic and cellular mechanisms remain poorly understood. We used a combination of in vitro slice electrophysiology, optogenetics, and in vivo two-photon imaging to investigate layer 5 extratelencephalic (ET) and layer 6 corticothalamic (CT) neuronal plasticity in mice, following noise-induced hearing loss (NIHL). Thalamocortical input was initially balanced between CTs and ETs but shifted to CT-dominant 1 day post-NIHL and then normalized by day 7. This transient shift was accompanied by increased quantal size and suprathreshold excitability in CTs, with minimal changes in ETs. In vivo, CTs exhibited persistent elevation in sound intensity thresholds, while ETs showed a transient shift in frequency tuning and reduced high-frequency responsiveness that recovered within a week. These findings reveal distinct, cell type–specific plasticity mechanisms in deep-layer ACtx neurons following peripheral damage and highlight potential targets for treating hearing loss–related disorders such as tinnitus and hyperacusis.

Noise-induced hearing loss triggers cell type–specific synaptic plasticity that reshapes cortical sound processing.

INTRODUCTION

In all sensory systems, damage to peripheral organs induces compensatory plasticity mechanisms in the brain, which contributes to preserving function by increasing neural responsiveness to surviving sensory inputs. In the cortex, this homeostatic plasticity is characterized by increased central gain and has been well documented across sensory modalities (1–5). In the auditory system, exposure to loud noise induces cochlear damage that often leads to a reduction in auditory nerve input due to cochlear synaptopathy and the degeneration or death of both hair and spiral ganglion cells (6–8). Despite this peripheral damage, the sound-evoked activity of excitatory cortical principal neurons is frequently restored, or even enhanced, within days of noise-induced hearing loss [NIHL; (9–13)], reflecting robust cortical and subcortical homeostatic plasticity mechanisms. These mechanisms support perceptual hearing threshold recovery, especially in quiet backgrounds, even after a marked loss of peripheral input (11, 14).

In the auditory cortex (ACtx), noise-induced cochlear damage is associated with enhanced central gain, reduced cortical inhibitory (GABAergic) activity, increased spontaneous firing, and reorganization of frequency tuning toward less damaged regions of the cochlea (11, 12, 15–20). Recent studies in ACtx layer (L) 2/3 neurons revealed that this plasticity is mediated by cell type–specific mechanisms, enabling a precise and cooperative division of labor among cortical interneurons (13). However, the plasticity mechanisms governing excitatory projection neurons remain less understood. Addressing this knowledge gap requires examining the distinct contributions of deep-layer neurons, specifically L5 extratelencephalic neurons (ETs, also known as pyramidal tract or PT neurons) and L6 corticothalamic neurons (CTs), as L5 and L6 represent the primary output pathways from the ACtx. CTs exclusively target the thalamus and readily modulate local cortical gain, while also modulating ongoing thalamocortical (TC) oscillatory activity, forming key nodes in cortico-thalamo-cortical loops (21–24). ETs also project to the thalamus as well as the midbrain, striatum, and amygdala (24–28). These projections are crucial for connecting the auditory system to downstream targets involved in motor control, sensorimotor integration, learning, and higher-order cognitive processes (29–34). It is therefore essential to determine how ETs, CTs, and their TC inputs change after NIHL. Given the importance of cortico-thalamo-cortical loops in perception and cognition in normal and pathological states (24, 35–39), understanding these mechanisms may highlight novel approaches to enhance perceptual recovery after hearing loss and mitigate plasticity-induced disorders such as tinnitus and hyperacusis (40).

To address these gaps in knowledge, we used a mouse model of NIHL to investigate plasticity mechanisms after noise trauma. We used a combination of in vitro slice electrophysiology and in vivo two-photon imaging to characterize changes in TC synaptic strength, intrinsic excitability, and sound response properties of ETs and CTs after NIHL. Our findings highlight cell type–specific plasticity mechanisms, offering insights into cortical plasticity in response to peripheral damage.

RESULTS

To induce a stereotyped hearing loss, we followed standard approaches (see Materials and Methods) and exposed mice to 8 to 16 kHz broadband noise at 116 dB SPL for 1 hour [noise exposed (NE)]. As a control, a separate cohort of littermate mice was sham exposed (SE) using identical procedures, but without noise presentation. To assess hearing thresholds pre- and post-NE or -SE [1 day before exposure (“−1d” hereinafter), 1 day after exposure (“1d” hereinafter), and 7 days after exposure (“7d” hereinafter)], we measured auditory brainstem responses (ABRs) (fig. S1, A to H), which reflect the synchronous activity of auditory nuclei, from the auditory nerve (wave I) to the inferior colliculus (IC; wave V) (fig. S1I). Following NE, we observed significantly elevated ABR thresholds and reduced wave I amplitudes both 1 day (fig. S1, A to I) and 7 days (fig. S2, A to H) after NE, across all tested frequencies. In contrast, ABR thresholds and wave I amplitudes were unchanged in SE mice (fig. S1, A to I). Together, these data confirm that our NE protocol produces a persistent hearing threshold shift lasting at least 1 week.

Transient shift in TC synaptic strength from CT and ET equivalent to CT dominant 1 day after NE

Communication between ACtx and the auditory thalamus, the medial geniculate body (MGB), is mediated by reciprocally connected corticothalamic (CT) and TC pathways. TC neurons in the MGB provide strong input to ACtx, innervating both ETs and CTs (41, 42). However, the impact of NIHL on TC synaptic strength onto these two projection neuron populations remains unknown.

To address this question, we performed dual recordings from ETs and CTs within the same brain slice while photoactivating TC axons. This recording configuration enabled us to calculate the ratio of synaptic strength at TC → CT versus TC → ET synapses (the CT/ET ratio) under identical experimental conditions. Although comparing absolute EPSC amplitudes with optogenetic recordings is not ideal due to variability in viral infection efficiency per mouse/brain area, the dual recording approach of the CT/ET ratio is well suited for detecting relative, noise-induced changes in TC input strength between CTs and ETs.

To study TC → CT and TC → ET synapses, we drove expression of channelrhodopsin-2 (ChR2) in CAMKII-expressing MGB neurons by injecting AAV-CaMKII-ChR2-YFP into the MGB (Fig. 1, A to C; see Materials and Methods). To stimulate ChR2-expressing TC axons, we used a collimated blue LED light source directed through a diaphragm and a 40× microscope objective lens. Monosynaptic photo-evoked AMPA receptor–mediated EPSCs were recorded simultaneously from vertically aligned ETs (green) and CTs (red) (Fig. 1, A to C). To selectively label ETs and delineate ACtx boundaries, we injected green fluorescent microspheres (retrobeads) into the right IC (Fig. 1, A and 1C, middle) (43, 44). To selectively label CTs, we injected Ntsr1-Cre transgenic mice, which selectively express Cre-recombinase in CTs, with AAV-FLEX-tdTomato (Fig. 1C, right). All electrophysiological experiments were performed in the presence of tetrodotoxin (TTX) to block action potentials (APs) and 4-aminopyridine (4-AP) to enhance presynaptic depolarization (see Materials and Methods). Because CT dendrites are primarily restricted to the deep cortical layers, photostimulation was confined to L5-L6 (Fig. 1B).

Fig. 1. Transient shift in TC synaptic strength from CT and ET equivalent to CT dominant 1 day after NE.

(A) Schematic illustration of stereotaxic injections of retrograde microspheres to label L5 extratelencephalic neurons (ETs, green), and viral vectors (AAVs) for expression of ChR2 in thalamocortical (TC) inputs and tdTomato in L6 corticothalamic neurons (CTs, red) in Ntsr1-Cre mice. (B) Schematic illustration of slice electrophysiology experiment involving photostimulation of ChR2 expressing TC afferents and simultaneous (dual) recording from an ET (green) and a CT (red). (C) Images in 4× magnification showing the extent of ACtx area in bright-field (left), green-labeled ETs and TC axons (middle), and red-labeled CTs. (D) Average CT/ET EPSC ratio after optogenetically stimulating thalamic L5-L6 input 1 day (1d) and 7 days (7d) after sham exposure (SE) and noise exposure (NE) (1d SE: 18 cells/6 mice; 1d NE: 26 cells/9 mice; 7d SE: 10 cells/6 mice; 7d NE: 10 cells/6 mice). Asterisks indicate significant differences (***P < 0.001, two-way ANOVA and Bonferroni corrected for multiple comparisons). (E and F) Representative traces of excitatory postsynaptic currents (EPSCs) in dual recordings from both CT (solid line) and ET (dotted line) neurons evoked by maximal photostimulation of L5-L6 thalamocortical inputs in 1d SE [(E), left] and 1d NE [(E), right]; 7d SE [(F), left] and 7d NE [(F), right]. Different colored traces represent different pairs of simultaneously recorded CTs and ETs. Detailed statistical values are listed in Table 1.

One day after NE, we observed a marked shift in TC input from equivalent (CT/ET ratio ~ 1) to CT dominant (CT/ET ratio > 5) [two-way analysis of variance (ANOVA), main effect for exposure, F = 5.44, P = 0.02; main effect for day, F = 2.25, P = 0.14]. By 7 days postexposure, TC input had recovered to a balanced state, with equivalent excitation of CTs and ETs (Fig. 1, D and F). Together, these results indicate that NIHL transiently enhances thalamic excitation of CTs relative to ETs, which recovers 7 days after NE.

Transient increase in the q on TC → CT synapses 1 day after NE

To understand the mechanisms underlying the shift toward greater relative thalamic excitation of CTs relative to ETs, we next examined the synaptic properties of TC → CT and TC → ET synapses, as well as the intrinsic excitability of CTs and ETs following NE. Because these recordings were performed in the presence of TTX and 4-AP, and the ChR2 activation directly depolarized presynaptic terminals, the probability of release (p) was likely saturated at all synapses (n). We therefore chose to assess the quantal size (q), first, in TC → CT synapses.

To evaluate the amplitude of quantal events, we replaced calcium (Ca²⁺) with strontium (Sr²⁺) in the bath solution (Fig. 2, A to D). Sr²⁺ desynchronizes evoked neurotransmitter release, allowing the analysis of quantal events originating from the stimulated synapses (Fig. 2, A to D) (45, 46). Following maximal photostimulation of L5-L6, we observed a significant increase in the average amplitude of quantal events at TC → CT synapses 1 day after NE (unpaired t test, P = 0.04; Fig. 2, A and B), which recovered by 7 days after NE (unpaired t test, P = 0.68; Fig. 2, C and D). This increase in q likely reflects enhanced synaptic strength at TC → CT synapses, either due to postsynaptic changes or changes in vesicle content. Consistent with these findings, we also found that average TC → CT EPSC amplitudes, elicited by L5-L6 photostimulation, were significantly enhanced 1 day after NE compared to SE (Mann-Whitney U test, P = 0.004; Fig. 2, E and F), which then recovered by 7 days (unpaired t test, P = 0.94; Fig. 2, G and H). Together, these results suggest that an increase in q at TC → CT synapses contribute to the greater relative thalamic excitation in TC → CT compared to TC → ET after NE.

Fig. 2. Transient increase in the q of TC→CT synapses 1 day after NE.

(A and C) Average amplitude of light-evoked quantal EPSCs (Sr²⁺-mEPSCs) in CTs in response to L5-L6 thalamocortical maximal photostimulation in 1 day (A) and 7 days (C), post-SE (black) and -NE (red) mice (1d SE: 7 cells/3 mice; 1d NE: 6 cells/3 mice; 7d SE: 5 cells/2 mice; 7d NE: 6 cells/2 mice). Asterisks indicate significant differences (*P < 0.05, unpaired t test). (B and D) Representative Sr²⁺-mEPSCs traces (top left) and representative average traces of quantal Sr²⁺-mEPSCs traces (top right). Amplitude histogram of events before (background noise) and after stimulation from the same cell (bottom) in 1 day (B) and 7 days (D) after SE (black) and NE (red) mice. The arrowhead indicates the onset of light stimulus. The dotted line indicates 400 ms time window before stimulus (pre-LED). The solid line represents a 400-ms time window, which started 100 ms after the stimulus (post-LED) and was used to analyze the amplitude of Sr²⁺-mEPSCs (1d SE: 7 cells/3 mice; 1d NE: 6 cells/3 mice; 7d SE: 5 cells/2 mice; 7d NE: 6 cells/2 mice). (E and G) Summary graph of the average CT EPSC amplitudes in response to L5-L6 thalamocortical maximal photostimulation in 1 day (E) and 7 days (G) SE (black) and NE (red) (1d SE: 18 cells/6 mice; 1d NE: 26 cells/9 mice; 7d SE: 10 cells/6 mice; 7d NE: 10 cells/6 mice). Asterisks indicate significant differences (**P < 0.01, Mann-Whitney U test). (F and H) Representative traces of L6 CT EPSCs in response to L5-L6 thalamocortical maximal photostimulation in 1d SE [(F), black line] and 1d NE [(F), red line] and 7d SE [(H), black line] and 7d NE [(H), red line]. Detailed statistical values are listed in Table 1.

Transient increase in CT suprathreshold intrinsic excitability 1 day after NE

Next, we investigated the effect of NE on CT intrinsic excitability. Neither input resistance (R_input) nor AP width or threshold (AP_width, AP_thres) changed at either 1 day or 7 days after NE (Fig. 3, A, C, and D). However, the resting membrane potential (V_rest) was significantly more hyperpolarized 1 day after NE (two-way ANOVA, main effect for exposure, F = 19.81, P < 0.0001; main effect for day, F = 9.40, P = 0.003; exposure × day interaction, F = 6.91, P = 0.01; Fig. 3B). We also observed a significant difference in the V_rest of SE mice between 1 and 7 days after exposure (two-way ANOVA, Bonferroni corrected for multiple comparisons, P = 0.0007). However, this difference does not affect the main conclusions of the study, which are based on comparisons between well-controlled SE and NE groups. When we evaluated suprathreshold excitability, we found that CTs displayed a significant increase in overall firing frequency in response to depolarization (two-way ANOVA, main effect for exposure, F = 9.45, P = 0.004; exposure × current steps interaction, F = 3.82, P < 0.0001; Fig. 3, E and F), suggesting an increased suprathreshold excitability of CTs concurrent with increased TC → CT input 1 day after NE. There were no differences in CT intrinsic suprathreshold excitability between SE and NE mice 7 days after exposure (two-way ANOVA, main effect for exposure, F = 2.64, P = 0.11; exposure × current steps interaction, F = 0.91, P = 0.52; Fig. 3, G and H). Together, these results suggest that despite a transient hyperpolarization of V_rest after NE, a combination of intrinsic and synaptic mechanisms that increase TC → CT synaptic strength and CT suprathreshold excitability likely contribute to the shift of TC input from equivalent between CT and ET to CT dominant 1 day after NE.

Fig. 3. Transient increase in suprathreshold CT intrinsic excitability 1 day after NE.

(A) Average CT input resistance (R_input) 1 and 7 days after SE (black) or NE (red) (1d SE: 22 cells/8 mice; 1d NE: 23 cells/5 mice; 7d SE: 23 cells/6 mice; 7d NE: 21 cells/4 mice). (B) Average CT resting membrane potential (V_rest) 1 and 7 days after SE (black) or NE (red) (1d SE: 22 cells/8 mice; 1d NE: 23 cells/5 mice; 7d SE: 23 cells/6 mice; 7d NE: 21 cells/4 mice). (C) Average CT AP width 1 and 7 days after SE (black) or NE (red) (1d SE: 21 cells/8 mice; 1d NE: 21 cells/5 mice; 7d SE: 23 cells/6 mice; 7d NE: 21 cells/4 mice). (D) Average CT AP threshold of 1 and 7 days after SE (black) or NE (red) (1d SE: 21 cells/8 mice; 1d NE: 19 cells/5 mice; 7d SE: 23 cells/6 mice; 7d NE: 21 cells/4 mice). (E and G) Representative traces of CT firing in response to depolarizing current (50, 100, and 200 pA current injections), 1d SE (E, left) versus 1d NE [(E), right], and 7d SE [(G), left] versus 7d NE [(G), right]. (F and H) Average firing frequency as a function of injected current amplitude, 1d SE versus 1d NE (F), and 7d SE versus 7d NE (H). Current injections from 25 to 250 pA with an increment of 25 pA (1d SE: 21 cells/8 mice; 1d NE: 21 cells/5 mice; 7d SE: 23 cells/6 mice; 7d NE: 21 cells/4 mice). Asterisks indicate significant differences (**P < 0.01, two-way ANOVA and Bonferroni corrected for multiple comparisons). Detailed statistical values are listed in Table 1.

No changes in the q of TC → ET synapses after NE

We next assessed TC → ET synapses using a similar electrophysiological approach as described for TC → CT synapses (Fig. 2). We found no significant changes in q at either 1 day or 7 days following NE (1d, Mann-Whitney U test, P = 0.64; 7d, Mann-Whitney U test, P > 0.9999; Fig. 4, A to D). However, the average TC → ET EPSC amplitudes were significantly reduced 1 day after NE compared to SE (two-way ANOVA, main effect for exposure, F = 8.32, P = 0.005; main effect for day, F = 0.05, P = 0.82; exposure × day interaction, F = 4.97, P = 0.03; Fig. 4, E and F). This reduction recovered by 7 days post-NE (Fig. 4, E and F). Given that p is likely saturated in TC → ET synapses, and q remained unchanged, the reduction in L5-L6 TC → ET EPSCs likely reflects a decrease in n. Alternatively, this result may be influenced by the inherent limitations of our experimental setup, particularly variability in AAV infection efficiency across mice and brain areas, which can complicate interpretation of synaptic strength (see Discussion).

Fig. 4. No changes in the q of TC→ET synapses after NE.

(A and C) Average amplitude of Sr²⁺-mEPSCs in ETs in response to L5-L6 thalamocortical maximal photostimulation in 1 day (A) and 7 days (C), post-SE (black) and -NE (red) (1d SE: 5 cells/4 mice; 1d NE: 7 cells/5 mice; 7d SE: 7 cells/2 mice; 7d NE: 5 cells/2 mice). (B and D) Representative Sr²⁺-mEPSCs traces (top left) and representative average traces of quantal Sr²⁺-mEPSCs traces (top right). Amplitude histogram of events before (background noise) and after stimulation from the same cell (bottom) in 1 day (B) and 7 days (D) after SE (black) and NE (red) mice. The arrowhead indicates the onset of light stimulus. The dotted line indicates a 400-ms time window before stimulus (pre-LED). The solid line represents a 400-ms time window, which started 100 ms after the stimulus (post-LED) and was used to analyze the amplitude of Sr²⁺-mEPSCs (1d SE: 5 cells/4 mice; 1d NE: 7 cells/5 mice; 7d SE: 7 cells/2 mice; 7d NE: 5 cells/2 mice). (E and G) Summary graph of the average L5 ET EPSC amplitudes in response to L5-L6, and L1-L4 thalamocortical maximal photostimulation in 1d SE versus 1d NE, or 7d SE versus 7d NE mice (L5-L6: 1d SE: 18 cells/6 mice; 1d NE: 26 cells/9 mice; 7d SE: 11 cells/6 mice; 7d NE: 10 cells/6 mice; L1-L4: 1d SE: 8 cells/3 mice; 1d NE: 7 cells/4 mice; 7d SE: 9 cells/5 mice; 7d NE: 12 cells/4 mice). (F and H) Representative traces of L5 ET EPSCs in response to L5-L6, and L1-L4 thalamocortical maximal photostimulation in 1d SE [(F) left, (H) left, black] versus 1d NE [(F) left, (H) left, red], and 7d SE [(F) right, (H) right, black] versus 7d NE [(F) right, (H) right, red] mice. Detailed statistical values are listed in Table 1.

Because ETs have extensive dendritic arbors throughout the cortical column, and TC projections to ETs span several cortical layers (25, 47), we also evaluated TC → ET EPSCs elicited after photostimulating L1-L4 (L1 -L4 TC EPSCs). No significant differences were observed at either 1 day or 7 days after NE (two-way ANOVA, main effect for exposure, F = 1.88, P = 0.18; main effect for day, F = 1.43, P = 0.24; exposure × day interaction, F = 3.35, P = 0.08; Fig. 4, G and H). Together, these findings support a cell type–specific increase in q at TC → CT synapses, which likely contributes to the enhanced thalamic excitation onto CTs relative to ETs 1 day after NE. In addition, our results suggest that the observed increase in the CT/ET ratio is likely influenced by the reduction in TC → ET EPSC amplitude (Fig. 4E), although the absolute magnitude of this decrease should be interpreted with caution due to the limitations of our optogenetic stimulation approach (see Discussion).

No changes in ET intrinsic excitability after NE

After establishing that NIHL differentially affects TC input onto ETs versus CTs, we next examined whether NE also alters the intrinsic excitability of these neurons in a cell type–specific manner. R_input, V_rest, AP_width, AP_threshold, and firing frequency-current (F-I) functions were unchanged at both 1 and 7 days post-NE in ETs (Fig. 5, A to H). This finding supports cell type–specific intrinsic plasticity in deep layer neurons after NIHL, whereby ETs remain intrinsically stable while CTs exhibit a transient decrease in subthreshold excitability and a transient increase in suprathreshold excitability.

Fig. 5. No changes in ET intrinsic excitability after NE.

(A) Average ET R_input 1 and 7 days after SE (black) or NE (red) (1d SE: 21 cells/6 mice; 1d NE: 21 cells/5 mice; 7d SE: 20 cells/5 mice; 7d NE: 23 cells/6 mice). (B) Average ET V_rest 1 and 7 days after SE (black) or NE (red) (1d SE: 21 cells/6 mice; 1d NE: 21 cells/5 mice; 7d SE: 20 cells/5 mice; 7d NE: 23 cells/6 mice). (C) Average ET AP width 1 and 7 days after SE (black) or NE (red) (1d SE: 21 cells/6 mice; 1d NE: 21 cells/5 mice; 7d SE: 20 cells/5 mice; 7d NE: 23 cells/6 mice). (D) Average ET AP threshold 1 and 7 days after SE (black) or NE (red) (1d SE: 21 cells/6 mice; 1d NE: 21 cells/5 mice; 7d SE: 20 cells/5 mice; 7d NE: 23 cells/6 mice). (E and G) Representative traces of ET firing in response to depolarizing current injection (50, 100, and 200 pA), 1d SE [(E), left] versus 1d NE [(E), right], and 7d SE [(G), left] versus 7d NE [(G), right]. (F and H) Average firing frequency as a function of injected current amplitude, 1d SE versus 1d NE (F), and 7d SE versus 7d NE (H). Current injections from 25 to 350 pA with an increment of 25 pA (1d SE: 21 cells/6 mice; 1d NE: 20 cells/5 mice; 7d SE: 20 cells/5 mice; 7d NE: 23 cells/6 mice). Detailed statistical values are listed in Table 1.

Changes in CT sound response properties persist 7 days after NE

To examine the consequences of NIHL in ET and CT sound response properties, we recorded the sound-evoked activity of neural populations using two-photon calcium imaging. This technique allowed us to longitudinally image the same populations of genetically defined cell types across time before and after noise exposure and characterize changes in sound response properties. Head-fixed, awake mice passively listened to auditory stimuli, and we recorded responses at −1d, then again at 1d and 7d, matching our previous electrophysiological studies. We selectively expressed GCaMP6s in CTs to examine how CT responses to sounds change following NE (Fig. 6A, see Materials and Methods). We imaged a total of 4943 CTs across 8 unique fields of view in SE mice (−1d: 1649 neurons; 1d: 1681 neurons; 7d: 1613 neurons) and 7731 CTs across 14 unique fields of view in NE mice (−1d: 2687 neurons; 1d: 2546 neurons; 7d: 2498 neurons). We then determined which neurons were significantly active during sound presentation and used these sound-responsive neurons for all subsequent analyses (see Materials and Methods).

Fig. 6. Changes in CT sound response responses persist 7 days after NE.

(A) Example two-photon field of view (FoV) from CTs imaged before and after NE. (B) Average pure tone responses across all CTs in SE mice (−1d: n = 507 neurons; 1d: n = 460 neurons; 7d: n = 432 neurons). (C) Same as (A) for NE mice (−1d: n = 713 neurons; 1d: n = 638 neurons; 7d: n = 646 neurons). (D) Tuning curves aligned to BF in SE mice. (E) Same as (D) for NE mice. (F) Change in BF of matched cells in SE (black) and NE (red) mice (1d SE: n = 40 neurons; 1d NE: n = 36 neurons; 7d SE: n = 35 neurons; 7d NE: n = 40 neurons). (G) Accuracy of a multinomial logistic regression classifier trained to decode pure tone frequency at 80 dB SPL. The dashed line represents chance level. Error bars represent SD of decoding iterations. (H) Average white noise responses at different intensities across all ETs in SE mice (−1d: n = 438 neurons; 1d: n = 460 neurons; 7d: n = 411 neurons). (I) Same as (H) in NE mice (−1d: n = 636 neurons; 1d: n = 559 neurons; 7d: n = 587 neurons). (J) Change in intensity thresholds across days in SE (black) and NE (red) mice (1d SE: n = 27 neurons; 1d NE: n = 31 neurons; 7d SE: n = 29 neurons; 7d NE: n = 25 neurons). (K) Accuracy of a multinomial logistic regression trained to decode sound intensity. The dashed line represents chance. Error bars represent SD of decoding iterations. All asterisks indicate significant pairwise differences (*P < 0.05, **P < 0.01, ***P < 0.001, Bonferroni corrected for multiple comparisons). Detailed statistical values are listed in Table 1.

We first examined frequency responses by presenting pure tones of varying frequencies (4 to 45 kHz in 0.5 octave steps) and intensities (20 to 80 dB SPL in 20 dB SPL steps; fig. S3B). Averaging across intensity, we initially observed that CTs exhibited unstable frequency encoding, as evidenced by fluctuations in average pure tone responses in SE animals (two-way ANOVA, main effect for day, F = 0.66, P = 0.52; frequency × day interaction, F = 3.68, P < 0.0001; Fig. 6B). Nevertheless, we observed strong changes in CT sound responses following NE (two-way ANOVA, main effect for day, F = 11.62, P < 0.0001; frequency × day interaction, F = 2.78, P = 0.0004; Fig. 6C). CTs exhibited increased population responses to low-frequency sounds (4 to 11 kHz) 7 days after NE. Next, we compared average population tuning curves by aligning activity to each neuron’s best frequency (BF) and comparing these aligned tuning curves across days. Frequency tuning curves remained stable in SE mice (two-way ANOVA, main effect for day, F = 0.52, P = 0.59; frequency × day interaction, F = 0.81, P = 0.67; Fig. 6D). NE induced a slight but significant increase in responses 0.5 octaves below BF (two-way ANOVA, main effect for day, F = 9.36, P < 0.0001; frequency × day interaction, F = 0.52, P = 0.94; Fig. 6E). To investigate whether individual neurons altered their tuning preferences following noise exposure, we matched the same cells across days and quantified changes in BF compared to baseline on day −1. While not significant, CTs displayed a trend to decrease BF following NE (two-way ANOVA, main effect for exposure, F = 1.00, P = 0.32; main effect for day, F = 1.14, P = 0.32; frequency × day interaction, F = 1.33, P = 0.27; Fig. 6F). To examine how these changes in frequency responses affected population encoding, we used a multinomial logistic regression model to predict stimulus frequency from neural activity. Decoding frequency identity from CT activity did not reveal any effect of noise exposure on population-level encoding (Fig. 6G). Together, these results suggest that noise exposure does not profoundly change CT frequency tuning.

We next examined intensity tuning by presenting broadband noise bursts between 20 and 80 dB SPL (in 10-dB SPL steps). As with pure tone responses, CTs showed subtle changes in broadband noise responses following SE (two-way ANOVA, main effect for day, F = 3.54, P = 0.03; frequency × day interaction, F = 2.51, P = 0.003; Fig. 6H). In contrast, there was a marked decrease in intensity responses following NE (two-way ANOVA, main effect for day, F = 13.87, P < 0.0001; frequency × day interaction, F = 5.45, P < 0.0001; Fig. 6I). Specifically, CTs exhibited decreased activity in response to low-intensity noise (<60 dB SPL) following NE, which did not recover by 7 days. There was also a slight but significant increase in responses to 80 dB SPL noise 7 days after NE. Consequently, NE led to a significant increase in intensity thresholds (two-way ANOVA, main effect for exposure, F = 14.75, P = 0.0002; main effect for day, F = 4.165, P = 0.02; frequency × day interaction, F = 4.94, P = 0.008; Fig. 6J). This was also evident when looking at the absolute intensity thresholds (fig. S4A). Despite these changes, decoding analyses did not reveal any significant changes in population encoding (Fig. 6K). Together, this suggests that noise exposure leads to decreased sensitivity in CTs to low-intensity noise that does not recover within a week.

Transient changes in ET sound response properties 1 day after NE

We next examined the activity of ETs through selective expression of GCaMP6s (Fig. 7A; see Materials and Methods). We imaged a total of 2337 ETs across nine unique fields of view in five SE mice (−1d: 705 neurons; 1d: 820 neurons; 7d: 812 neurons) and 1703 ET neurons across eight unique fields of view in four NE mice (−1d: 562 neurons; 1d: 565 neurons; 7d: 576 neurons). As before, only neurons that were significantly active during sound presentation were used for subsequent analyses.

Fig. 7. Transient shift in ET sound response properties 1 day after NE.

(A) Example two-photon FoV from ETs imaged before and after NE. (B) Average pure tone responses across all ETs in SE mice (−1d: n = 385 neurons; 1d: n = 414 neurons; 7d: n = 400 neurons). (C) Same as in (A) for NE mice (−1d: n = 268 neurons; 1d: n = 235 neurons; 7d: n = 296 neurons). (D) Tuning curves aligned to best frequency (BF) in SE mice. (E) Same as in (D) for NE mice. (F) Change in BF of matched cells in SE (black) and NE (red) mice (1d SE: n = 68 neurons; 1d NE: n = 54 neurons; 7d SE: n = 68 neurons; 7d NE: n = 71 neurons). (G) Accuracy of a multinomial logistic regression classifier trained to decode pure tone frequency at 80 dB SPL. The dashed line represents chance. Error bars represent SD of decoding iterations. (H) Average white noise responses at different intensities across all ETs in SE mice (−1d: n = 331 neurons; 1d: n = 270 neurons; 7d: n = 299 neurons). (I) Same as (H) in NE mice (−1d: n = 226 neurons; 1d: n = 222 neurons; 7d: n = 252 neurons). (J) Change in intensity thresholds across days in SE (black) and NE (red) mice (1d SE: n = 42 neurons; 1d NE: n = 49 neurons; 7d SE: n = 50 neurons; 7d NE: n = 51 neurons). (K) Accuracy of a multinomial logistic regression trained to decode sound intensity. The dashed line represents chance. Error bars represent SD of decoding iterations. All asterisks indicate significant pairwise differences (*P < 0.05, **P < 0.01, ****P < 0.00001, Bonferroni corrected for multiple comparisons). Detailed statistical values are listed in Table 1.

We found that the average population responses to pure tones remained unchanged in SE mice (two-way ANOVA, main effect for day, F = 0.68, P = 0.51; frequency × day interaction, F = 1.03, P = 0.41; Fig. 7B). In contrast, NE led to a shift in the representation of frequencies across the ET population (two-way ANOVA, main effect for day, F = 19.35, P < 0.0001; frequency × day interaction, F = 1.93, P = 0.02; Fig. 7C). One day after NE, the average response to 23-kHz pure tones significantly decreased compared to baseline. This reduction in activity then recovered to baseline levels by 7 days and was accompanied by a significant increase in response to lower-frequency sounds (11 to 16 kHz). BF-aligned tuning curves were consistent across days in SE mice (two-way ANOVA, main effect for day, F = 0.58, P = 0.56; frequency × day interaction, F = 0.58, P = 0.90; Fig. 7D). Conversely, NE led to significant changes in the average tuning curves (two-way ANOVA, main effect for day, F = 12.03, P < 0.0001; frequency × day interaction, F = 1.84, P = 0.02; Fig. 7E). Specifically, 7 days after NE, ETs increased the magnitude of their responses at, and one octave below, BF. This was accompanied by a significant decrease in BF itself on day 1, which then recovered by day 7 in NE animals (two-way ANOVA, main effect for exposure, F = 17.67, P < 0.0001; main effect for day, F = 2.23, P = 0.11; exposure × day interaction, F = 11.34, P < 0.0001; Fig. 7F). Frequency decoding revealed that SE populations exhibited stable predictive accuracy across days, while NE led to a significant decrease in accuracy 1 day after NE and a subsequent recovery 7 days after NE (Fig. 7G). These findings suggest that 1 day following NE, ETs reduce their activity at high frequencies and consequently shift to lower BFs, decreasing the ability to accurately decode stimuli from population activity. These perturbations then return to baseline 7 days after NE.

We next examined whether ET intensity tuning changed following NE. As expected, noise responses were similar across days in SE mice (two-way ANOVA, main effect for day, F = 0.20, P = 0.82; intensity × day interaction, F = 0.94, P = 0.50; Fig. 7H). We observed that NE led to significant changes in intensity responses (two-way ANOVA, main effect for day, F = 8.01, P = 0.0003; frequency × day interaction, F = 2.90, P = 0.0005; Fig. 7I). One day after NE, ETs displayed a loss of responses to low intensities (<50 dB SPL) and a reduction in responses at higher intensities. After 7 days, responses partially recovered but remained below baseline for 70 dB SPL. We then quantified changes in single neuron intensity thresholds with cells matched across imaging days (two-way ANOVA, main effect for exposure, F = 11.23, P = 0.0009; main effect for day, F = 6.56, P = 0.002; exposure × day interaction, F = 3.496, P = 0.03; Fig. 7J). One day after NE, ETs displayed an increased threshold, which reverted to baseline by 7 days. This recovery was also evident when looking at the absolute intensity thresholds (fig. S4D). We also observed a threshold decrease 7 days after SE, driven by a slightly higher threshold before exposure. Last, we decoded sound intensity from population activity using multinomial logistic regression. Similar to our frequency decoding findings, we found that decoding accuracy transiently decreased 1 day following NE (Fig. 7K). Together, these findings suggest that ETs recover from transient changes in tuning to both frequency and intensity following noise exposure.

Changes in synaptic, intrinsic, and sound response properties correlate with hearing loss severity

To address whether the degree of neural plasticity was tied to the extent of hearing loss or recovery in individual animals, we correlated each animal’s ABR click threshold shift with the magnitude of the synaptic, intrinsic, and sound response changes. We found that many of the measured plastic changes scaled significantly with trauma severity (Fig. 8). In particular, the relative strengthening of TC input to CTs was more pronounced in animals with greater hearing loss: The CT/ET EPSC ratio was positively correlated with ABR threshold shift (Fig. 8A; Spearman’s ρ = 0.46, P = 0.002). The average L5-L6 TC → CT q and EPSC amplitude increased with worsening ABR thresholds (Fig. 8, B and C; ρ = 0.34, P = 0.03), whereas TC → ET EPSCs were inversely related to hearing status (Fig. 8D; ρ = −0.36, P = 0.02), further supporting both an enhanced synaptic TC drive onto CTs and a diminished TC drive onto ETs in mice with more severe hearing loss. In line with these synaptic effects, the intrinsic properties of CTs also covaried with hearing loss: Animals with larger ABR threshold shifts had more hyperpolarized CT resting potentials (Fig. 8E; ρ = −0.51, P = 0.002). We also noted that the increase in CT firing showed an overall positive trend with ABR shift (Fig. 8F; maximum firing frequency; ρ = 0.31, P = 0.08), although this did not reach significance, likely due to the limited sample size for this measure, which is a general limitation for these correlations. Alternatively, this result could suggest that suprathreshold excitability changes occurred uniformly after trauma rather than scaling with severity of hearing loss.

Fig. 8. Changes in synaptic, intrinsic, and sound response properties correlate with noise trauma severity.

(A) Correlation between ABR threshold shift and CT/ET EPSC ratio after optogenetically stimulating thalamic input (Fig. 1D) in 1d SE (black) and NE (red) mice (1d SE: 18 cells/6 mice; 1d NE: 26 cells/9 mice). The line depicts a linear fit. (B) Same as (A) for amplitude of light-evoked quantal EPSCs in CTs in response to TC photostimulation (Fig. 2A) in 1d SE (black) and NE (red) mice (1d SE: 7 cells/3 mice; 1d NE: 6 cells/3 mice). (C) Same as (A) for average CT EPSC amplitudes in response to TC photostimulation (Fig. 2E) in 1d SE (black) and NE (red) mice (1d SE: 18 cells/6 mice; 1d NE: 26 cells/9 mice). (D) Same as (A) for ET EPSC amplitudes in response to TC photostimulation (Fig. 4E) in 1d SE (black) and NE (red) mice (1d SE: 18 cells/6 mice; 1d NE: 26 cells/9 mice). (E) Same as (A) for CT resting membrane potential (Fig. 3B) in 1d SE (black) and NE (red) mice (1d SE: 22 cells/8 mice; 1d NE: 23 cells/5 mice). (F) Same as (A) for the maximum firing frequency (Fig. 3F) in 1d SE (black) and NE (red) mice (1d SE: 21 cells/8 mice; 1d NE: 21 cells/5 mice). (G and H) Same as (A) for the change in intensity thresholds (Fig. 6J) in 1d (G) or 7d (H) SE (black) and NE (red) mice (1d SE: n = 27 neurons; 1d NE: n = 31 neurons; 7d SE: n = 29 neurons; 7d NE: n = 25 neurons). (I and J) Same as (A) for the change in intensity thresholds (Fig. 7J) in 1d (I) or 7d (J) SE (black) and NE (red) mice (1d SE: n = 42 neurons; 1d NE: n = 49 neurons; 7d SE: n = 50 neurons; 7d NE: n = 51 neurons).

We next asked whether the in vivo sound responses were related to the degree of hearing loss or recovery across animals. One day after exposure, the elevation in CT neuronal threshold (Δ intensity threshold) was positively correlated with ABR shift (Fig. 8G; ρ = 0.32, P = 0.01). Animals with persistently high ABR thresholds retained larger CT threshold elevations 7 days postexposure (Fig. 8H; ρ = 0.31, P = 0.02). The transient threshold shift observed in ETs 1 day after exposure also scaled with hearing loss (Fig. 8I; ρ = 0.23, P = 0.03). In contrast, ETs no longer showed a correlation by 7 days (Fig. 8J; ρ = 0.07, P = 0.47), consistent with their response properties largely recovering despite the sustained peripheral deficit. Thus, interanimal variability in both synaptic and sensory response plasticity generally tracked the severity of NIHL, whereas the rapid recovery of the ET projection by 7 days occurred regardless of residual hearing impairment.

DISCUSSION

Our findings provide insights into the cellular and circuit mechanisms underlying the cortical plasticity of deep-layer neurons in the ACtx following NIHL. We identified distinct changes in the synaptic, intrinsic, and sound response properties of L5 extratelencephalic (ET) and L6 CT neurons. One day after NIHL, TC input was temporarily reweighted to favor CTs over ETs. We hypothesize that this transient shift in relative thalamic drive acts as a compensatory mechanism to preserve auditory signaling following peripheral damage. Overall, the cell type–specific plasticity observed in ETs and CTs advances our understanding of how ACtx adapts to cochlear injury and preserves hearing function post-NIHL.

Enhanced TC input and intrinsic excitability in L6 CT neurons after NIHL

CTs form reciprocal and nonreciprocal feedback loops with the thalamus to modulate TC activity through gain control and temporal filtering (21–23, 48–50). Optogenetic experiments revealed a selective enhancement of TC input to CTs 1 day after NE. In simultaneous recordings of neighboring CTs and ETs, CTs received significantly stronger excitation from thalamic afferents compared to ETs 1 day after NE (Fig. 1). Although our optogenetic approach could be influenced by variability in viral infection and a likely saturation of release probability under our recording conditions (Figs. 2, E and H, and 4, E and H), the within-animal comparison between CTs and ETs effectively controls for the variability in viral infection. Quantal analysis, which permits comparison of synaptic strength across different animals and brain slices, also showed an increased quantal amplitude (q) at TC synapses onto CTs (Fig. 2, A to D), supporting a genuine up-regulation of TC quantal synaptic strength onto CT circuits after noise trauma.

In parallel with these synaptic changes, CTs became intrinsically more excitable 1 day after NIHL, especially for suprathreshold stimuli (Fig. 3, E and F). The combination of enhanced synaptic excitation and intrinsic excitability suggests that CTs are poised to fire more readily in response to thalamic input following trauma. Such rapid strengthening of CT excitability likely represents an early homeostatic adjustment aimed at preserving cortical activity levels despite reduced cochlear input. Because CTs provide feedback to both the lemniscal (primary) and nonlemniscal auditory thalamus, their increased excitability could boost the overall gain of TC transmission and stabilize network activity across multiple circuits. This fast CT up-regulation may be a compensatory mechanism to enhance perceptual processing after NIHL, potentially through improved sensory gain control and/or multisensory/contextual modulation (37, 39, 51–53).

Persistent reduction in low-intensity sound responses of L6 CT neurons

Despite the overall increase in excitability of CTs, we observed a paradoxical reduction in their responsiveness to low-intensity sounds following NIHL. One day after NIHL, CTs responded weakly to near-threshold tones, as shown by sound-evoked measurements in vivo (Fig. 6, H to K). This loss of sensitivity to faint sounds persisted 1 week after NIHL and cannot be explained by the enhanced TC input or intrinsic excitability described above (which would predict an increase in firing). One potential contributing factor is the slight hyperpolarization of CT V_rest observed 1 day after NIHL (Fig. 3B), which could raise the spiking threshold. However, this intrinsic change is likely not the main cause, because by 7 days post-NIHL, V_rest had returned to baseline, yet their low-intensity sound responses remained reduced. It is possible that changes in intracortical excitation or inhibition onto CTs may be selectively gating out weaker inputs after trauma. It is possible, for example, that CTs received enhanced feedforward inhibition or that their spike timing was altered in vivo in a way that requires a stronger synchronized input to elicit firing. Although we did not directly measure changes in L6 inhibitory circuits, any increase in local inhibition would further limit CT activity in response to weak stimuli, counterbalancing their higher intrinsic and thalamus-driven excitability.

Why might ACtx suppress responses to low-intensity sounds after NIHL? One interpretation is that the auditory system is trading off raw sensitivity for reliability. By raising the threshold for ACtx responses to soft inputs, the system could avoid spurious firing to unreliable background noise. This idea is consistent with the notion that CT feedback circuits help resolve competing demands of sound detection versus discrimination (21–23). This nuanced plasticity in CTs, whereby global excitability increases but responsiveness to weak inputs decreases, potentially optimizes the signal-to-noise ratio of auditory cortical output after NIHL.

Minimal TC and intrinsic changes with transient retuning in L5 ET neurons after NIHL

In contrast to CTs, ETs exhibited relatively minimal changes in their intrinsic and synaptic properties after NIHL. We observed no significant alterations in ETs’ intrinsic excitability at either 1 day or 7 days post-NIHL (Fig. 5), consistent with previous studies (54).

While the quantal amplitude (q) of TC synapses onto ETs remained unaltered (Fig. 4, A to D), the most notable early synaptic effect on ETs was a transient reduction in overall thalamic drive observed 1 day post-NIHL (Fig. 4, E and F). Optogenetic EPSCs recorded from ETs were smaller at 1 day post-NIHL, which might indicate a decrease in the number of functional thalamic release sites (n), given the likely saturated probability of release during our recordings. We acknowledge that our optical stimulation approach has limitations (including variability in viral transduction and possible saturation of release probability); thus, the absolute magnitude of this reduction should be interpreted cautiously. Nevertheless, by 7 days post-NIHL, the TC input to ETs had recovered (Fig. 4, E and F), and the balance of thalamic drive between CTs and ETs returned to near pre-NIHL levels. The early drop in ET input is thus best viewed as a short-term adjustment, likely contributing to the shift we saw toward CT-dominant thalamic excitation 1 day after noise trauma. Even at its weakest point, some thalamic input to ETs was preserved, allowing these neurons to continue participating in cortical processing during the acute phase after NIHL.

Although ETs showed little change in their excitability, they did undergo a rapid shift in sound-evoked tuning after NIHL. One day post-NIHL, ETs became less responsive to high-frequency tones (Fig. 7). To compensate for this loss, ETs increased their gain 7 days after NE. As a result, ETs recovered their frequency tuning (Fig. 7F) but conversely responded proportionally more to lower-frequency inputs that were spared by the trauma (Fig. 7, C and E). Furthermore, ET population encoding largely returned to baseline (Fig. 7, G and K), coinciding with the recovery of TC input strength to ETs (Fig. 4, E and F). Such transient retuning and lasting gain enhancement likely helps the animal continue to perceive sounds in the frequency ranges that remain intact, thereby partially compensating for the loss at damaged frequencies.

Given the minimal intrinsic and postsynaptic changes in ETs, this reorganization is likely mediated by enhanced excitatory and/or decreased inhibitory intracortical inputs; however, our studies did not study these inputs. ETs are major sources of corticofugal projections to subcortical structures; thus, maintaining their sound-evoked activity may ensure continued engagement and/or trigger plasticity in downstream targets like the MGB and the IC during the acute posttrauma period. This is supported by related findings that corticofugal outputs from L5 neurons become hyperactive after cochlear injury (55). This sustained potentiation is thought to underlie network hyperexcitability and reorganization in the aftermath of hearing loss. Thus, the transient bias of ETs toward lower-frequency responsiveness in our study may reflect a functional routing of cortical drive to preserve auditory throughput to the thalamus and midbrain.

A caveat in our approach is that we identified ETs and CTs using different labeling strategies (retrobead tracing versus Cre-dependent viral tagging). Retrobeads remain sequestered in vesicles and are designed to be inert, causing minimal if any impact on neuronal health (56). Similarly, AAV labeling yields long-term fluorescent protein expression with low toxicity and negligible effect on neuronal physiology at the doses used. Both approaches are widely used, often together, and are not known to alter intrinsic excitability or synaptic function (42–44, 57–59). Moreover, any subtle labeling-related effects (e.g., mild immune activation from viral infection) would be expected to affect SE and NE groups equally and are therefore unlikely to confound the main conclusions of our study.

Layer-specific adaptive strategies in ACtx after NIHL

Our study of deep-layer ETs and CTs complements and extends recent research on plasticity in superficial layers of ACtx (11–14), revealing a layer-specific pattern of adaptation after NIHL. Recent results showed that L2/3 interneurons exhibit cell type–specific plastic changes following NIHL, with different subclasses of inhibitory neurons showing distinct time courses and levels of recovery (13, 14). Such findings underscore that each cortical layer (and cell type) may deploy unique mechanisms to cope with degraded sensory input. In superficial layers, the priority may be maintaining local network stability and gain control. Adjustments in L2/3 circuits (especially among inhibitory interneurons) likely serve to enhance principal neuron gain while preventing hyperexcitability when thalamic input is reduced or imbalanced (13). Although our study focused on plasticity in the deeper layers, we did not directly examine GABAergic interneuron changes. However, it is likely that the infragranular adaptations we observed are modulated by the alterations in inhibition occurring in supragranular layers. One potential mechanism is through cortical disinhibition: Supragranular vasoactive intestinal peptide interneurons, which become hyperactive posttrauma (13), inhibit parvalbumin (PV) interneurons across layers (60–62), including those that normally target the apical dendrites of infragranular pyramidal cells. Thus, a reduction in PV activity after trauma would release deep-layer neurons from dendritic inhibition, allowing them to respond more strongly.

On the other hand, ETs and CTs engage in plasticity that appears geared toward preserving output and driving broader compensatory changes. These projection neurons are the main conduits of cortical influence on subcortical structures (ETs projecting to the thalamus, midbrain, and other subcortical centers, and CTs projecting exclusively to the thalamus). The plastic changes in deep layers, such as enhanced CT excitability and transient ET retuning, suggest that deep-layer circuits are working to propagate adaptive signals throughout the auditory pathway and perhaps trigger plasticity in lemniscal, nonlemniscal, and even multisensory pathways. For instance, the rapid CT plasticity might initiate compensatory adjustments in thalamic relay neurons or even in other cortical areas by altering the feedback they provide to the MGB. Similarly, stable firing in ETs ensures that activity continues to flow to multiple subcortical centers, which could drive plastic reweighting or gain changes.

Complementary roles of ET and CT plasticity after NIHL

Viewed together, the divergent changes in ETs and CTs represent complementary adaptations that serve the shared goal of maintaining auditory function after NIHL. CTs, with their strengthened thalamic inputs and heightened intrinsic excitability, primarily contribute to normalizing cortical activity and refining sensory processing. By increasing the gain of TC transmission, CTs amplify the cortical representation of incoming sounds, which can compensate for reduced peripheral input. At the same time, our data suggest that CT circuits impose a stricter threshold for low-intensity inputs thereby filtering out noise and enhancing the discrimination of important signals. This dual action of CT plasticity (boosting overall responsiveness while gating insignificant inputs) aligns with models in which cortical feedback optimizes the trade-off between detection and discrimination (21–23). In practical terms, the CT pathway may be elevating the “signal” and suppressing the “noise”: It reduces spurious activity to trivial stimuli (or spontaneous activity from a damaged cochlea) and ensures that meaningful auditory cues still evoke strong cortical responses. Such a mechanism would preserve perceptual stability and potentially improve functions like speech understanding in noisy environments.

ETs likely ensure that the overall drive to subcortical auditory structures is maintained. Even though ETs did not markedly change their own excitability, their transient shift to favor responsiveness to lower-frequency sounds and the putative recruitment of intracortical inputs allow them to continue sending signals to downstream nuclei. By preserving ET activity, the ACtx guarantees that key downstream centers remain engaged and active during the critical period after injury. ET circuits provide a broad corticofugal output and are crucial for high-level cognitive processes, acoustic-motor behaviors, and auditory plasticity (30, 63–65). Thus, this sustained subcortical drive could itself trigger adaptive plasticity in those centers.

It should be noted that although these mechanisms are aimed at preserving hearing function after cochlear damage, they operate on a delicate balance. Even within ostensibly homeostatic/compensatory changes, there is a risk of overcompensation, hyperactivity and TC dysrhythmia, which can contribute to disorders like tinnitus and hyperacusis (39, 40, 66)

In conclusion, our work reveals cell type–specific plasticity in ACtx ETs and CTs following noise trauma. The observed changes in synaptic strength, intrinsic properties, and sound response tuning suggest a coordinated adaptive response aimed at preserving auditory function directly or via triggering adaptive plasticity in MGB and the ACtx. These findings contribute to a broader understanding of cortical plasticity in sensory processing and may inform therapeutic strategies for enhancing recovery from hearing loss and mitigating hearing loss–related disorders, such as tinnitus and hyperacusis.

MATERIALS AND METHODS

Animals

All mice handling was approved by the Institutional Animal Care and Use Committee at the University of Pittsburgh. A total of 66 male and 38 female Ntsr1-Cre mice [MMRRC, B6.FVB(Cg)-Tg(Ntsr1-Cre)GN220Gsat/Mmcd] were used for experiments shown in Figs. 1 to 5 and figs. S1 and S2; 3 male and 1 female Ntsr1-Cre mice as well as 4 male and 5 female Ntsr1-Cre × Ai148 mice [JAX, Ai148(TIT2L-GC6f-ICL-tTA2)-D] were used for Fig. 6; 7 male and 2 female C57Bl/6J mice were used for Fig. 7.

Surgical methods for in vitro electrophysiology

Mice (P28 to P35) were anesthetized with isoflurane (induction: 3% in oxygen, maintenance: 1.5 to 2% in oxygen) and secured in a stereotaxic frame (Kopf). Core body temperature was maintained at ~37°C with a heating pad and eyes were protected with ophthalmic ointment. Lidocaine (1%) was injected under the scalp and a midline incision was made to expose the skull. ETs were retrogradely labeled by injecting green fluorescent latex microspheres (Lumafluor) into the right IC (1 mm posterior to lambda and 1 mm lateral, at an injection depth of 0.75 mm), schematized in Fig. 1A. To transduce CTs, an AAV (AAV9-FLEX-tdTomato, titer: 1 × 10¹³ gc/ml, 1:2 dilution in PBS, Addgene) was injected into the right ACtx of Ntsr1-Cre mice through a small burr hole (~0.4 mm diameter, ~4 mm lateral to lambda). A glass micropipette containing the viral vector was lowered into ACtx to a depth of ~1 mm using a micromanipulator (Kopf). A syringe pump (World Precision Instruments) was then used to inject 200 nl over the course of 2 min, at a rate of 100 nl/min.

To facilitate stimulation of thalamic afferents, 500 nl of AAV9-CaMKIIa-hChR2(H134R)-EYFP (titer: 8.96 × 10¹³ gc/ml, Addgene) was injected into the right MGB of Ntsr1-Cre mice (3 mm posterior/2 mm lateral/3 mm deep) at a rate of 100 nl/min, alongside injections of microspheres and AAV9-FLEX-tdTomato, as described earlier. This resulted in widespread expression across multiple MGB subdivisions. Postinjection, the scalp was closed with cyanoacrylate adhesive and mice were given a subcutaneous injection of carprofen (5 mg/kg).

Surgical methods for in vivo two-photon imaging

Mice (P56 to P72) were anesthetized with isoflurane (induction: 5% in oxygen, maintenance: 1.5 to 2% in oxygen) and secured in a stereotaxic frame (Kopf). Dexamethasone (2 mg/kg) was administered via an intraperitoneal injection to mitigate brain edema. Core body temperature was maintained at ~37°C with a heating pad and eyes were protected with ophthalmic ointment. Lidocaine (1%) was injected under the scalp and a midline incision was made to expose the skull. To transduce ETs, an AAV (AAVretro-hSyn1-GCaMP6s-P2A-nls-tdTomato, titer: 5 × 10¹² gc/ml, Addgene) was injected into the right IC of C57 mice through a small burr hole (5 mm posterior to bregma and 1.2 mm lateral). A glass micropipette containing the viral vector was lowered to depths of 0.3 and 0.8 mm using a micromanipulator (Kopf). A syringe pump (Nanoject III, Drummond) was used to inject 300 nl at each site at a rate of 10 to 15 nl/min. CTs were transduced with GCaMP6s through intracranial injection into the ACtx (AAV5-Syn-FLEX-GCaMP6s-WPRE-SV40, titer: 7 × 10¹² gc/ml, Addgene) or through transgenic crossing with the Ai148 mouse line. Postinjection, the scalp was closed with nylon sutures and mice were administered with a subcutaneous injection of carprofen (5 mg/kg). Viruses were allowed to express for a minimum of 3 weeks before two-photon imaging. Cranial windows were created by stacking three round glass coverslips (one 4 mm, two 3 mm, #1 thickness, Warner Instruments), etched with piranha solution (sulfuric acid and hydrogen peroxide), then glued together with optically transparent, ultraviolet-cured adhesive (Norland Products). The skull was exposed by removing the dorsal portion of the scalp and periosteum, then treated with etchant (C&B metabond) followed by 70% ethanol. A custom stainless steel headplate (iMaterialize, (67) was then attached to the skull using dental cement (C&B metabond). A 3-mm circular craniotomy was made over right ACtx and the cranial window was placed into the opening. An airtight seal was achieved using Kwik-Sil (World Precision Instruments) around the edge of the window before cementing it into place. Vetbond (3M) was used to bind the surrounding tissue to the dental cement. Following the implant surgery, mice were given a subcutaneous injection of carprofen (5 mg/kg). Implants were allowed to heal for at least 5 days before the first day of imaging.

Slice electrophysiology

Mice (P58 to P65) were first anesthetized with isoflurane and then immediately decapitated. Brains were rapidly removed and coronal slices (300 μm) containing right ACtx were prepared in a cutting solution at 1°C using a vibratome (VT1200 S, Leica). The cutting solution (pH 7.4, ∼300 mOsm) contained the following: 2.5 mM KCl, 1.25 mM NaH₂PO₄, 25 mM NaHCO₃, 0.5 mM CaCl₂, 7 mM MgCl₂, 7 mM glucose, 205 mM sucrose, 1.3 mM ascorbic acid, and 3 mM sodium pyruvate (bubbled with 95% O₂/5% CO₂). Slices were immediately transferred to a holding chamber and incubated at 34°C for 40 min before recording. The holding chamber contained artificial cerebrospinal fluid (ACSF), with the following composition: 125 mM NaCl, 2.5 mM KCl, 26.25 mM NaHCO₃, 2 mM CaCl₂, 1 mM MgCl₂, 10 mM glucose, 1.3 mM ascorbic acid, and 3 mM sodium pyruvate, pH 7.4, ∼300 mOsm (bubbled with 95% O₂/5% CO₂). After incubation, slices were stored at room temperature until recording.

Whole-cell recordings in voltage- and current-clamp modes were performed in slices bathed in 31°C carbogenated ACSF, identical to the incubating solution. Borosilicate pipettes (World Precision Instruments) were pulled into patch electrodes with a resistance of 3 to 6 megohms (Sutter Instruments). For current-clamp recordings, pipettes were filled with a potassium-based intracellular solution, containing the following: 128 mM K-gluconate, 10 mM Hepes, 4 mM MgCl₂, 4 mM Na₂ATP, 0.3 mM tris-GTP, 10 mM tris phosphocreatine, 1 mM EGTA, and 3 mM sodium ascorbate (pH 7.25, 295 mOsm). For voltage-clamp recordings of optogenetically evoked excitatory postsynaptic currents (EPSCs), a cesium-based internal solution was used, containing the following: 126 mM CsCH₃O₃S, 4 mM MgCl₂ 10 mM Hepes, 4 mM Na₂ATP, 0.3 mM tris-GTP, 10 mM tris-phosphocreatine, 1 mM CsEGTA, 1 mM QX-314, and 3 mM sodium ascorbate (pH 7.25, 295 mOsm). Recordings were made using a MultiClamp-700B amplifier with a Digidata-1440A A/D converter and Clampex software (Molecular Devices). Data were sampled at 10 kHz and filtered at 4 kHz. Pipette capacitance was compensated, and series resistance (R_series) was maintained below 25 megohms. R_series was measure in voltage clamp mode (command potential set at −70 mV) by applying a −5-mV voltage step and calculating the ratio of the step voltage to the peak current immediately following the step. Input resistance (R_input) was calculated in voltage-clamp mode (command potential set to −70 mV) by giving a −5-mV step, which resulted in transient current responses. The difference between baseline and steady-state hyperpolarized current (ΔI) was used to calculate R_input using the following formula: R_input = −5 mV/ΔI – R_series. Resting membrane potential (V_rest) was measured in current-clamp mode (I = 0) by averaging membrane potential for 20 s after break-in. For intrinsic property measurements, the following drugs were included in the bath: 20 μΜ DNQX (AMPA receptor antagonist), 50 μΜ APV (NMDA receptor antagonist), and 20 μΜ SR 95531 hydrobromide (Gabazine—a GABA_A receptor antagonist). Depolarizing current pulses (25-pA increments of 1 s duration) were injected from a holding potential of −70 mV (adjusted with direct current if necessary) to assess suprathreshold firing properties. AP width was defined as the full width at the half-maximum amplitude of the first AP at rheobase. AP threshold was determined from the phase-plane plot as the membrane potential where the depolarization slope (dV/dt) exceeded 10 V/s.

Light-evoked EPSCs and strontium quantal events (Sr²⁺-mEPSCs) were evoked by optogenetic stimulation of presynaptic axons. A collimated blue LED light source (470 nm, Thorlabs) was directed through a diaphragm and a 40× microscope objective lens, restricting illumination to a small spot within the targeted cortical layer. ETs and CTs were selected for dual whole-cell recordings based on visual alignment along the radial (depth) axis using differential interference contrast optics. To ensure anatomical proximity consistent with columnar organization, we assessed the tangential (horizontal) distance between the somata of the recorded pairs. Post hoc measurements across all recorded pairs revealed that the mean tangential distance was approximately 20 μm (18.81 ± 1.86 μm; n = 59 pairs). The light intensity required to elicit a stable, maximal plateau response was determined for each cell pair using a 0.15-ms light pulse. EPSCs were recorded in voltage clamp mode at −70 mV. Peak amplitudes were averaged over a 1-ms window. To isolate monosynaptic responses, recordings were performed in the presence of TTX (1 μM) to block sodium channels and 4-AP (100 μM) potassium channel blocker to depolarize the terminals. Representative traces for figures were generated by averaging 10 consecutive EPSCs.

Sr²⁺-mEPSCs were recorded in slices incubated for 30 min in a modified Sr²⁺-ACSF solution containing the following: 125 mM NaCl, 2.5 mM KCl, 26 mM NaHCO₃, 4 mM SrCl₂, 4 mM MgCl₂, 15 mM glucose, 1.3 mM ascorbic acid, and 3 mM sodium pyruvate, pH 7.4, ∼300 mOsm, oxygenated with 95% O₂-5% CO₂. All other recording conditions, including the cutting solution, cesium-based internal solution, and light stimulation, remained the same as described above. Data were analyzed using Clampfit (11.2). A 400-ms window preceding LED onset (pre-LED) and a 400-ms window beginning 100 ms after LED onset (post-LED) were used for event detection. The amplitude of LED-evoked quantal events was calculated using the equation: $\frac{[(A_{\text{post}}\ast F_{\text{post}})-(A_{\text{pre}}\ast F_{\text{pre}})]}{F_{\text{post}}-F_{\text{pre}}}$ , where A_post is the average amplitude of post-LED Sr²⁺-mEPSCs, F_post is the frequency of post-LED events, A_pre is the average amplitude of pre-LED events, and F_pre is their frequency (68). Event detection was optimized by low-pass filtering the traces with a 1-kHz Gaussian filter. Events were included if they had a rise time <3 ms and amplitude >3 × root mean square (RMS) noise. Cells with RMS noise >2 or where F_post − F_pre < 2 Hz were excluded from the analysis.

Noise and sham exposure

Twenty-three to twenty-nine after viral injections, mice were anesthetized with isoflurane (induction: 3% in oxygen, maintenance: 1.5 to 2% in oxygen) during a unilateral noise exposure procedure. A plastic pipette tip, connected to a calibrated speaker (FT17H, Fostex), was inserted into the left ear. Core body temperature was maintained at ~37°C with a heating pad and eyes were protected with ophthalmic ointment. SE and NE mice were treated identically, but NE mice were presented with 8 to 16 kHz broadband noise at 116 dB SPL for 1 hour.

Bandwidth limiting was achieved using a DS360 Ultra-Low Distortion Function Generator (Stanford Research Systems) in combination with a 3-pole Butterworth filter. The speaker (Type 4231, Brüel & Kjær) was calibrated using a 1/4-inch (6.35-mm) microphone (Model 4954-B, Brüel & Kjær) to accurately determine the output voltage required to achieve the desired sound level.

Auditory brainstem responses

ABR measurements were made from the left (exposed) ear and conducted in a sound-attenuating chamber (ENV-022SD, Med Associates). Subdermal electrodes were positioned with the active electrode at the vertex, the ground electrode ventral to the right pinna, and the reference electrode ventral to the left pinna. A calibrated speaker (CF-1, Tucker-Davis Technologies) delivered acoustic stimuli to the left ear via a pipette tip affixed to a plastic tube connected to the speaker.

ABR measurements were recorded in response to 1-ms tone bursts at 10, 16, 24, and 32 kHz, presented in descending order from 80 to 0 dB SPL (or until 10 dB SPL below threshold). Threshold was defined as the lowest stimulus intensity that evoked a detectable wave 1 response. A wave 1 response was identified as the first consistent deflection that decreased in amplitude and increased in latency as sound intensity decreased, distinguishing it from background noise. ABR measurements were acquired using BioSigRX software at a rate of 18.56 Hz, averaged over 512 repetitions, and filtered with a 300- to 3000-Hz bandpass filter.

Two-photon imaging

Mice (P55 to P104) were acclimated to head fixation in the two-photon imaging rig for 30 min/day for three consecutive days before the start of experiments. Imaging was performed at three time points: −1d, 1d, and 7d. No additional head fixation occurred between 1d and 7d. ABRs were performed before each imaging session to assess peripheral sound responses and hearing status. Mice with ABR click thresholds of 50 dB SPL or greater on −1d were excluded from the study. NE mice were excluded if their click thresholds increased by less than 10 dB SPL from −1d to 1d. Similarly, SE mice were excluded if their click thresholds increased by more than 10 dB SPL across the same time points.

On −1d, widefield epifluorescence imaging was used to generate an ACtx tonotopic map, allowing identification of primary auditory area A1 (67). The two-photon field of view (FoV) was centered over A1 for all subsequent recordings. Two-photon excitation was provided by an Insight X3 laser tuned to 940 nm (Spectra-Physics). A 16×/0.8 numerical aperture water-immersion objective (Nikon) was used to obtain a 512 pixel by 512 pixel FoV at 30 Hz with a Bergamo II Galvo-Resonant 8-kHz scanning microscope (Thorlabs). Scanning software (Thorlabs) and stimulus generation hardware (National Instruments) were synchronized with digital pulse trains. For both widefield and two-photon imaging, the microscope was rotated 40° to 55° from the vertical axis to image the lateral portion of mouse ACtx while the mouse remained upright. Imaging was conducted in a dark, sound-attenuating chamber, and mice were monitored using infrared cameras (Genie Nano, Teledyne).

ETs were imaged 420 to 580 μm below the pial surface using laser powers of 45 to 130 mW, while CTs were imaged 600 to 790 μm below the pial surface with powers ranging from 111 to 150 mW. Fluorescence images were acquired at either 1× (750 μm by 750 μm) or 2× (375 μm by 375 μm) digital zoom. Raw calcium imaging data were processed using Suite2P (69) to perform rigid/nonrigid registration, region of interest (ROI) detection, and spike deconvolution. All ROIs were visually curated, and any low-quality or nonsomatic ROIs were excluded from further analysis. The same FoV was imaged across all three time points. Individual neurons were matched across days using ROIMatchPub, an open-source MATLAB plugin (https://github.com/ransona/ROIMatchPub).

Each imaging session included the presentation of two stimulus sets. Eight pure tones (4 to 45 kHz in half-octave steps) were presented at four intensities (20 to 80 dB SPL in 20-dB SPL steps). In addition, white noise bursts were presented at seven intensities (20 to 80 dB SPL in 10-dB SPL steps). All stimuli were 50 ms in duration, followed by a 2-s intertrial interval. Stimuli were delivered in a pseudorandom order, and each stimulus was repeated 20 times.

Two-photon imaging analysis

Deconvolved calcium traces (fig. S3A) were normalized for each cell by z-scoring relative to the average baseline activity during the 500 ms preceding sound onset. Evoked stimulus responses were quantified as the mean z-scored activity during the 0- to 500-ms window following sound onset. To determine whether a neuron was responsive to the given stimulus, the z-scored activity during the stimulus window was compared to 1000 samples of time-shifted responses, generated by circularly shifting the trace by 5 to 20 s. A neuron was considered significantly responsive if its evoked response exceeded the 98th percentile of the randomly shifted distribution (70). The BF of each neuron was defined as the frequency that evoked the highest z-scored response after averaging across intensities. Intensity thresholds were defined as the lowest sound intensity that produced a significant z-scored response.

Stimulus decoding

We trained a multinomial logistic regression classifier to decode stimulus identity (either pure tone frequency at 80 dB SPL or white noise intensity) on a trial-by-trial basis. All neurons were pooled across mice for each exposure and day condition. Using 5-fold cross-validation, the classifier was trained on the mean evoked activity (0 to 500 ms following sound onset) from 200 randomly selected neurons. Decoding accuracy was quantified as the fraction of correctly classified trials in the held-out test set. Because the full dataset contained between 562 and 2687 neurons per condition, we repeated this decoding procedure 2000 times, each time resampling (with replacement) a new population of 200 neurons. This resampling strategy ensures that decoding accuracy estimates were not biased by group size differences. We then calculated the mean and SD of the resultant decoding accuracy distribution for each exposure/day condition. P values were computed explicitly from the bootstrapped distributions of decoding accuracies (71).

Image processing

Image captioning for Fig. 1C was performed using Qcapture (QImaging). All images were imported and processed using ImageJ (https://imagej.nih.gov/ij/).

Statistics

For statistical comparisons between two independent groups that passed the Shapiro-Wilk test for normality, we used unpaired t tests. For nonnormally distributed data, Mann-Whitney U (rank-sum) tests were used. Comparisons involving more than two groups were assessed using N-way ANOVAs, with appropriate corrections for multiple comparisons. For correlation analyses, we calculated Spearman’s rank correlation coefficient (rho). Significance levels are denoted as *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. All group data are presented as mean ± SEM unless otherwise specified. See Table 1 and table S1 for detailed statistical values and test results.

Table 1. Statistical values.

For Figs. 1 to 7. −1d, 1 day before exposure; 1d, 1 day after exposure; 7d, 7 days after exposure.

Figure	Statistical test	Main effect 1	Main effect 2	Interaction effect	N (cells per mice)	Multiple comparison
1D	Two-way ANOVA Bonferroni	Exposure:	Day:		1d SE: 18/6	1d SE versus 1d NE, P = 0.0002
		F(1,60) = 5.44	F(1,60) = 2.25	F(1,60) = 7.19	1d NE: 26/9	7d SE versus 7d NE, P > 0.99
		P = 0.023	P = 0.14	P = 0.0095	7d SE: 10/6	1d SE versus 7d SE, P > 0.99
					7d NE: 10/6	1d NE versus 7d NE, P = 0.021
2A	Unpaired t test	n/a	n/a	n/a	1d SE: 7/3	P = 0.037
					1d NE: 6/3
2C	Unpaired t test	n/a	n/a	n/a	7d SE: 5/2	P = 0.68
					7d NE: 6/2
2E	Mann-Whitney U test	n/a	n/a	n/a	1d SE: 18/6	P = 0.0042
					1d NE: 26/9
2G	Unpaired t test	n/a	n/a	n/a	7d SE: 10/6	P = 0.94
					7d NE: 10/6
3A	Two-way ANOVA Bonferroni	Exposure:	Day:		1d SE: 22/8	ns for all groups
		F(1,85) = 0.15	F(1,85) = 2.26	F(1,85) = 0.33	1d NE: 23/5
		P = 0.70	P = 0.14	P = 0.57	7d SE: 23/6
					7d NE: 21/4
3B	Two-way ANOVA Bonferroni	Exposure:	Day:		1d SE: 22/8	1d SE versus 1d NE, P < 0.0001
		F(1,85) = 19.81	F(1,85) = 9.40	F(1,85) = 6.91	1d NE: 23/5	1d SE versus 7d SE, P = 0.0007
		P < 0.0001	P = 0.0029	P = 0.010	7d SE: 23/6	1d NE versus 7d NE, P > 0.99
					7d NE: 21/4	7d SE versus 7d NE, P > 0.99
3C	Two-way ANOVA Bonferroni	Exposure:	Day:		1d SE: 21/8	1d SE versus 1d NE, P > 0.99
		F(1,82) = 3.852	F(1,82) = 4.345	F(1,82) = 0.0081	1d NE: 21/5	1d SE versus 7d SE, P = 0.95
		P = 0.0531	P = 0.0402	P = 0.9284	7d SE: 23/6	1d NE versus 7d NE, P = 0.79
					7d NE: 21/4	7d SE versus 7d NE, P = 0.88
3D	Two-way ANOVA Bonferroni	Exposure:	Day:		1d SE: 21/8	1d SE versus 1d NE, P = 0.23
		F(1,80) = 3.73	F(1,80) = 0.91	F(1,80) = 1.26	1d NE: 19/5	1d SE versus 7d SE, P > 0.99
		P = 0.057	P = 0.34	P = 0.27	7d SE: 23/6	1d NE versus 7d NE, P = 0.93
					7d NE: 21/4	7d SE versus 7d NE, P > 0.99
3F	Two-way ANOVA Bonferroni	Exposure:	Current steps:		1d SE: 21/8	225 pA, P = 0.029 250 pA, P = 0.024 ns for all other groups
		F(1,40) = 9.45	F(2.431,97.24) = 155.5	F(10,400) = 3.82	1d NE: 21/5
		P = 0.0038	P < 0.0001	P < 0.0001
3H	Two-way ANOVA Bonferroni	Exposure:	Current steps:		7d SE: 23/6	ns for all groups
		F(1,42) = 2.64	F(2.369,99.49) = 208.3	F(10,420) = 0.91	7d NE: 21/4
		P = 0.11	P < 0.0001	P = 0.52
4A	Mann-Whitney U test	n/a	n/a	n/a	1d SE: 5/4	P = 0.64
					1d NE: 7/5
4C	Mann-Whitney U test	n/a	n/a	n/a	7d SE: 7/2	P > 0.99
					7d NE: 5/2
4E	Two-way ANOVA Bonferroni	Exposure:	Day:		1d SE: 18/6	1d SE versus 1d NE, P = 0.0002
		F(1,61) = 8.32	F(1,61) = 0.052	F(1,61) = 4.97	1d NE: 26/9	1d SE versus 7d SE, P > 0.99
		P = 0.0054	P = 0.82	P = 0.030	7d SE: 11/6	1d NE versus 7d NE, P = 0.50
					7d NE: 10/6	7d SE versus 7d NE, P > 0.99
4G	Two-way ANOVA Bonferroni	Exposure:	Day:		1d SE: 8/3	1d SE versus 1d NE, P > 0.99
		F(1,32) = 1.88	F(1,32) = 1.43	F(1,32) = 3.35	1d NE: 7/4	1d SE versus 7d SE, P > 0.99
		P = 0.18	P = 0.24	P = 0.077	7d SE: 9/5	1d NE versus 7d NE, P = 0.23
					7d NE: 12/4	7d SE versus 7d NE, P = 0.12
5A	Two-way ANOVA Bonferroni	Exposure:	Day:		1d SE: 21/6	ns for all groups
		F(1,81) = 0.27	F(1,81) = 0.016	F(1,81) = 1.41	1d NE: 21/5
		P = 0.60	P = 0.90	P = 0.24	7d SE: 20/5
					7d NE: 23/6
5B	Two-way ANOVA Bonferroni	Exposure:	Day:		1d SE: 21/6	ns for all groups
		F(1,81) = 1.64	F(1,81) = 0.002	F(1,81) = 1.09	1d NE: 21/5
		P = 0.20	P = 0.96	P = 0.30	7d SE: 20/5
					7d NE: 23/6
5C	Two-way ANOVA Bonferroni	Exposure:	Day:		1d SE: 21/6	ns for all groups
		F(1,81) = 3.80	F(1,81) = 0.014	F(1,81) = 0.23	1d NE: 21/5
		P = 0.055	P = 0.91	P = 0.63	7d SE: 20/5
					7d NE: 23/6
5D	Two-way ANOVA Bonferroni	Exposure:	Day:		1d SE: 21/6	ns for all groups
		F(1,81) = 1.60	F(1,81) = 0.12	F(1,81) = 0.64	1d NE: 21/5
		P = 0.21	P = 0.73	P = 0.43	7d SE: 20/5
					7d NE: 23/6
5F	Two-way ANOVA Bonferroni	Exposure:	Current steps:		1d SE: 21/6	ns for all groups
		F(1,39) = 3.386	F(2.016,78.63) = 815.9	F(14,546) = 1.274	1d NE: 20/5
		P = 0.073	P < 0.0001	P = 0.22
5H	Two-way ANOVA Bonferroni	Exposure:	Current steps:		7d SE: 20/5	ns for all groups
		F(1,41) = 0.022	F(2.604,106.8) = 716	F(14,574) = 1.07	7d NE: 23/6
		P = 0.88	P < 0.0001	P = 0.39
6B	Two-way ANOVA Bonferroni					4 kHz: −1d versus 1d, P = 0.18; −1d versus 7d, P = 0.52; 1d versus 7d, P > 0.99
						6 kHz: −1d versus 1d, P = 0.016; −1d versus 7d, P = 0.14; 1d versus 7d, P > 0.99
		Frequency:	Day:		−1d: 507/5	8 kHz: −1d versus 1d, P = 0.40; −1d versus 7d, P = 0.20; 1d versus 7d, P > 0.99
		F(7,11,168) = 2.07	F(2,11,168) = 0.66	F(14,11,168) = 3.68	1d: 460/5	11 kHz: −1d versus 1d, P > 0.99; −1d versus 7d, P = 0.84; 1d versus 7d, P > 0.99
		P = 0.043	P = 0.52	P < 0.0001	7d: 432/5	16 kHz: −1d versus 1d, P > 0.99; −1d versus 7d, P > 0.99; 1d versus 7d, P > 0.99
						23 kHz: −1d versus 1d, P = 0.21; −1d versus 7d, P = 0.47; 1d versus 7d, P > 0.99
						32 kHz: −1d versus 1d, P = 0.031; −1d versus 7d, P = 0.0052; 1d versus 7d, P > 0.99
						45 kHz: −1d versus 1d, P < 0.0001; −1d versus 7d, P = 0.023; 1d versus 7d, P = 0.29
6C	Two-way ANOVA Bonferroni					4 kHz: −1d versus 1d, P > 0.99; −1d versus 7d, P = 0.0029; 1d versus 7d, P = 0.0064
						6 kHz: −1d versus 1d, P = 0.50; −1d versus 7d, P = 0.029; 1d versus 7d, P = 0.73
		Frequency:	Day:		−1d: 713/8	8 kHz: −1d versus 1d, P > 0.99; −1d versus 7d, P = 0.028; 1d versus 7d, P = 0.16
		F(7,15,952) = 19.68	F(2,15,952) = 11.62	F(14,15,952) = 2.78	1d: 638/8	11 kHz: −1d versus 1d, P = 0.081; −1d versus 7d, P = 0.0041; 1d versus 7d, P < 0.0001
		P < 0.0001	P < 0.0001	P = 0.0004	7d: 646/8	16 kHz: −1d versus 1d, P > 0.99; −1d versus 7d, P > 0.99; 1d versus 7d, P > 0.99
						23 kHz: −1d versus 1d, P = 0.46; −1d versus 7d, P = 0.26; 1d versus 7d, P > 0.99
						32 kHz: −1d versus 1d, P > 0.99; −1d versus 7d, P > 0.99; 1d versus 7d, P > 0.99
						45 kHz: −1d versus 1d, P > 0.99; −1d versus 7d, P = 0.41; 1d versus 7d, P > 0.99
6D	Two-way ANOVA	Frequency:	Day:		−1d: 507/5	n/a
		F(8,8780) = 195.0	F(2,8780) = 0.52	F(16,8780) = 0.81	1d: 460/5
		P < 0.0001	P = 0.59	P = 0.67	7d: 432/5
6E	Two-way ANOVA Bonferroni					BF − 2: −1d versus 1d, P > 0.99; −1d versus 7d, P > 0.99; 1d versus 7d, P > 0.99
						BF − 1.5: −1d versus 1d, P > 0.99; −1d versus 7d, P > 0.99; 1d versus 7d, P > 0.99
						BF − 1: −1d versus 1d, P > 0.99; −1d versus 7d, P = 0.43; 1d versus 7d, P = 0.17
		Frequency:	Day:		−1d: 713/8	BF − 0.5: −1d versus 1d, P = 0.67; −1d versus 7d, P = 0.03; 1d versus 7d, P = 0.0007
		F(8,12,876) = 166.4	F(2,12,876) = 9.36	F(16,12,876) = 0.52	1d: 638/8	BF: −1d versus 1d, P > 0.99; −1d versus 7d, P = 0.17; 1d versus 7d, P = 0.17
		P < 0.0001	P < 0.0001	P = 0.94	7d: 646/8	BF + 0.5: −1d versus 1d, P > 0.99; −1d versus 7d, P = 0.58; 1d versus 7d, P = 0.34
						BF+ 1: −1d versus 1d, P = 0.97; −1d versus 7d, P > 0.99; 1d versus 7d, P = 0.62
						BF + 1.5: −1d versus 1d, P > 0.99; −1d versus 7d, P = 0.65; 1d versus 7d, P > 0.99
						BF + 2: −1d versus 1d, P > 0.99; −1d versus 7d, P = 0.87; 1d versus 7d, P = 0.87
6F	Two-way ANOVA				−1d SE: 72/5	n/a
		Exposure:	Day:		1d SE: 40/5
		F(1,316) = 1.00	F(2,316) = 1.14	F(2,316) = 1.33	7d SE: 35/5
		P = 0.32	P = 0.32	P = 0.27	−1d NE: 99/8
					1d NE: 36/8
					7d NE: 40/8
6G	Bootstrap Bonferroni (see Materials and Methods)	n/a	n/a	n/a		−1d SE versus −1d NE: P = 0.63
					−1d SE: 507/5	1d SE versus 1d NE: P = 0.71
					1d SE: 460/5	7d SE versus 7d NE: P = 0.40
					7d SE: 432/5	−1d SE versus 1d SE: P > 0.99
					−1d NE: 713/8	−1d SE versus 7d SE: P > 0.99
					1d NE: 638/8	1d SE versus 7d SE: P > 0.99
					7d NE: 646/8	−1d NE versus 1d NE: P > 0.99
						−1d NE versus 7d NE: P > 0.99
						1d NE versus 7d NE: P > 0.99
6H	Two-way ANOVA Bonferroni					20 dB: −1d versus 1d, P > 0.99; −1d versus 7d, P = 0.90; 1d versus 7d, P > 0.99
						30 dB: −1d versus 1d, P > 0.99; −1d versus 7d, P > 0.99; 1d versus 7d, P > 0.99
		Intensity:	Day:		−1d: 438/5	40 dB: −1d versus 1d, P > 0.99; −1d versus 7d, P > 0.99; 1d versus 7d, P > 0.99
		F(6,9142) = 96.27	F(2,9142) = 3.54	F(12,9142) = 2.51	1d: 460/5	50 dB: −1d versus 1d, P = 0.38; −1d versus 7d, P = 0.55; 1d versus 7d, P = 0.013
		P < 0.0001	P = 0.029	P = 0.0027	7d: 411/5	60 dB: −1d versus 1d, P > 0.99; −1d versus 7d, P > 0.99; 1d versus 7d, P > 0.99
						70 dB: −1d versus 1d, P = 0.48; −1d versus 7d, P = 0.24; 1d versus 7d, P > 0.99
						80 dB: −1d versus 1d, P = 0.0014; −1d versus 7d, P < 0.0001; 1d versus 7d, P = 0.61
6I	Two-way ANOVA Bonferroni					20 dB: −1d versus 1d, P > 0.99; −1d versus 7d, P > 0.99; 1d versus 7d, P > 0.99
						30 dB: −1d versus 1d, P = 0.76; −1d versus 7d, P = 0.77; 1d versus 7d, P > 0.99
		Intensity:	Day:		−1d: 636/8	40 dB: −1d versus 1d, P = 0.0008; −1d versus 7d, P = 0.0001; 1d versus 7d, P > 0.99
		F(6,12,453) = 104.6	F(2,12,453) = 13.87	F(12,12,453) = 5.45	1d: 559/8	50 dB: −1d versus 1d, P < 0.0001; −1d versus 7d, P < 0.0001; 1d versus 7d, P > 0.99
		P < 0.0001	P < 0.0001	P < 0.0001	7d: 587/8	60 dB: −1d versus 1d, P = 0.50; −1d versus 7d, P = 0.021; 1d versus 7d, P = 0.63
						70 dB: −1d versus 1d, P > 0.99; −1d versus 7d, P = 0.69; 1d versus 7d, P = 0.48
						80 dB: −1d versus 1d, P > 0.99; −1d versus 7d, P = 0.026; 1d versus 7d, P = 0.0068
6J	Two-way ANOVA Bonferroni					−1d SE versus −1d NE: P > 0.99
					−1d SE: 66/5	1d SE versus 1d NE: P = 0.0064
					1d SE: 27/5	7d SE versus 7d NE: P = 0.0016
		Exposure:	Day:		7d SE: 29/5	−1d SE versus 1d SE: P = 0.27
		F(1,243) = 14.75	F(2,243) = 4.17	F(2,243) = 4.94	−1d NE: 71/8	−1d SE versus 7d SE: P > 0.99
		P = 0.0002	P = 0.017	P = 0.0079	1d NE: 31/8	1d SE versus 7d SE: P = 0.45
					7d NE: 25/8	−1d NE versus 1d NE: P = 0.36
						−1d NE versus 7d NE: P = 0.0006
						1d NE versus 7d NE: P = 0.14
6K	Bootstrap Bonferroni (see Materials and Methods)	n/a	n/a	n/a		−1d SE versus −1d NE: P = 0.85
					−1d SE: 438/5	1d SE versus 1d NE: P = 0.38
					1d SE: 460/5	7d SE versus 7d NE: P = 0.62
					7d SE: 411/5	−1d SE versus 1d SE: P > 0.99
					−1d NE: 636/8	−1d SE versus 7d SE: P > 0.99
					1d NE: 559/8	1d SE versus 7d SE: P > 0.99
					7d NE: 587/8	−1d NE versus 1d NE: P > 0.99
						−1d NE versus 7d NE: P > 0.99
						1d NE versus 7d NE: P > 0.99
7B	Two-way ANOVA	Frequency:	Day:		−1d: 385/5	n/a
		F(7,9568) = 18.96	F(2,9568) = 0.68	F(14,9568) = 1.04	1d: 414/5
		P < 0.0001	P = 0.51	P = 0.41	7d: 400/5
7C	Two-way ANOVA Bonferroni					4 kHz: −1d versus 1d, P > 0.99; −1d versus 7d, P = 0.087; 1d versus 7d, P = 0.098
						6 kHz: −1d versus 1d, P > 0.99; −1d versus 7d, P = 0.22; 1d versus 7d, P = 0.23
		Frequency:	Day:		−1d: 268/4	8 kHz: −1d versus 1d, P > 0.99; −1d versus 7d, P = 0.17; 1d versus 7d, P = 0.13
		F(7,6368) = 23.07	F(2,6368) = 19.35	F(14,6368) = 1.93	1d: 235/4	11 kHz: −1d versus 1d, P = 0.50; −1d versus 7d, P = 0.0003; 1d versus 7d, P = 0.054
		P < 0.0001	P < 0.0001	P = 0.0191	7d: 296/4	16 kHz: −1d versus 1d, P > 0.99; −1d versus 7d, P = 0.039; 1d versus 7d, P = 0.0055
						23 kHz: −1d versus 1d, P = 0.020; −1d versus 7d, P = 0.70; 1d versus 7d, P = 0.0003
						32 kHz: −1d versus 1d, P = 0.19; −1d versus 7d, P > 0.99; 1d versus 7d, P = 0.38
						45 kHz: −1d versus 1d, P = 0.33; −1d versus 7d, P = 0.43; 1d versus 7d, P > 0.99
7D	Two-way ANOVA	Frequency:	Day:		−1d: 385/5	n/a
		F(8,7831) = 99.52	F(2,7831) = 0.58	F(16,7831) = 0.58	1d: 414/5
		P < 0.0001	P = 0.56	P = 0.90	7d: 400/5
7E	Two-way ANOVA Bonferroni					BF − 2: −1d versus 1d, P > 0.99; −1d versus 7d, P > 0.99; 1d versus 7d, P > 0.99
						BF − 1.5: −1d versus 1d, P > 0.99; −1d versus 7d, P > 0.99; 1d versus 7d, P > 0.99
						BF − 1: −1d versus 1d, P > 0.99; −1d versus 7d, P = 0.031; 1d versus 7d, P = 0.0064
		Frequency:	Day:		−1d: 268/4	BF − 0.5: −1d versus 1d, P = 0.78; −1d versus 7d, P > 0.99; 1d versus 7d, P = 0.12
		F(8,5363) = 98.43	F(2,5363) = 12.03	F(16,5363) = 1.84	1d: 235/4	BF: −1d versus 1d, P = 0.54; −1d versus 7d, P < 0.0001; 1d versus 7d, P < 0.0001
		P < 0.0001	P < 0.0001	P = 0.021	7d: 296/4	BF + 0.5: −1d versus 1d, P > 0.99; −1d versus 7d, P = 0.96; 1d versus 7d, P = 0.18
						BF + 1: −1d versus 1d, P > 0.99; −1d versus 7d, P > 0.99; 1d versus 7d, P = 0.49
						BF + 1.5: −1d versus 1d, P > 0.99; −1d versus 7d, P > 0.99; 1d versus 7d, P > 0.99
						BF + 2: −1d versus 1d, P > 0.99; −1d versus 7d, P > 0.99; 1d versus 7d, P > 0.99
7F	Two-way ANOVA Bonferroni					−1d SE versus −1d NE: P > 0.99
					−1d SE: 114/5	1d SE versus 1d NE: P < 0.0001
					1d SE: 68/5	7d SE versus 7d NE: P = 0.40
		Exposure:	Day:		7d SE: 68/5	−1d SE versus 1d SE: P = 0.14
		F(1,480) = 17.67	F(2,480) = 2.23	F(2,480) = 11.34	−1d NE: 111/4	−1d SE versus 7d SE: P > 0.99
		P < 0.0001	P = 0.11	P < 0.0001	1d NE: 54/4	1d SE versus 7d SE: P = 0.067
					7d NE: 71/4	−1d NE versus 1d NE: P < 0.0001
						−1d NE versus 7d NE: P = 0.39
						1d NE versus 7d NE: P = 0.015
7G	Bootstrap Bonferroni (see Materials and Methods)	n/a	n/a	n/a		−1d SE versus −1d NE: P = 0.098
					−1d SE: 385/5	1d SE versus 1d NE: P < 0.0001
					1d SE: 414/5	7d SE versus 7d NE: P = 0.1750
					7d SE: 400/5	−1d SE versus 1d SE: P > 0.9999
					−1d NE: 268/4	−1d SE versus 7d SE: P > 0.9999
					1d NE: 235/4	1d SE versus 7d SE: P > 0.9999
					7d NE: 296/4	−1d NE versus 1d NE: P > 0.9999
						−1d NE versus 7d NE: P > 0.9999
						1d NE versus 7d NE: P = 0.0360
7H	Two-way ANOVA	Intensity:	Day:		−1d: 331/5	n/a
		F(6,6279) = 44.38	F(2,6279) = 0.2007	F(12,6279) = 0.9433	1d: 270/5
		P < 0.0001	P = 0.8182	P = 0.5019	7d: 299/5
7I	Two-way ANOVA Bonferroni					20 dB: −1d versus 1d, P > 0.9999; −1d versus 7d, P > 0.9999; 1d versus 7d, P > 0.9999
						30 dB: −1d versus 1d, P = 0.6319; −1d versus 7d, P = 0.6348; 1d versus 7d, P > 0.9999
		Intensity:	Day:		−1d: 226/4	40 dB: −1d versus 1d, P = 0.1788; −1d versus 7d, P = 0.0191; 1d versus 7d, P < 0.0001
		F(6,4879) = 47.49	F(2,4879) = 8.014	F(12,4879) = 2.900	1d: 222/4	50 dB: −1d versus 1d, P > 0.9999; −1d versus 7d, P = 0.0541; 1d versus 7d, P = 0.1216
		P < 0.0001	P = 0.0003	P = 0.0005	7d: 252/4	60 dB: −1d versus 1d, P = 0.0061; −1d versus 7d, P = 0.1869; 1d versus 7d, P = 0.5673
						70 dB: −1d versus 1d, P = 0.1120; −1d versus 7d, P = 0.0510; 1d versus 7d, P > 0.9999
						80 dB: −1d versus 1d, P = 0.5423; −1d versus 7d, P > 0.9999; 1d versus 7d, P = 0.2016
7J	Two-way ANOVA Bonferroni					−1d SE versus −1d NE: P > 0.9999
					−1d SE: 99/5	1d SE versus 1d NE: P = 0.0043
					1d SE: 42/5	7d SE versus 7d NE: P = 0.0148
		Exposure:	Day:		7d SE: 50/5	−1d SE versus 1d SE: P > 0.9999
		F(1,368) = 11.23	F(2,368) = 6.555	F(2,368) = 3.496	−1d NE: 83/4	−1d SE versus 7d SE: P = 0.0096
		P = 0.0009	P = 0.0016	P = 0.0313	1d NE: 49/4	1d SE versus 7d SE: P = 0.0806
					7d NE: 51/4	−1d NE versus 1d NE: P = 0.0068
						−1d NE versus 7d NE: P > 0.9999
						1d NE versus 7d NE: P = 0.0117
7K	Bootstrap Bonferroni (see Materials and Methods)	n/a	n/a	n/a		−1d SE versus −1d NE: P = 0.1970
					−1d SE: 331/5	1d SE versus 1d NE: P = 0.0450
					1d SE: 270/5	7d SE versus 7d NE: P = 0.0910
					7d SE: 299/5	−1d SE versus 1d SE: P > 0.9999
					−1d NE: 226/4	−1d SE versus 7d SE: P > 0.9999
					1d NE: 222/4	1d SE versus 7d SE: P > 0.9999
					7d NE: 252/4	−1d NE versus 1d NE: P > 0.9999
						−1d NE versus 7d NE: P > 0.9999
						1d NE versus 7d NE: P = 0.4230

Supplementary Materials

This PDF file includes:

Figs. S1 to S4

Table S1